Shubhankar Mohapatra

I am senior scientist at A*STAR Singapore in ASTAR Centre for Frontier AI Research (ACFAR) working on privacy and safety of AI systems. I received my PhD from the DSG lab at the Cheriton School of Computer Science with Prof. Xi He. I was also a Vector Institute graduate student and part of The Salon led by Prof. Gautam Kamath. My research interests lie in privacy-preserving learning algorithms with a focus on differential privacy, generative AI and private data cleaning.

I have worked as a research intern at Sony AI in the PPML team. Before UWaterloo, I had the opportunity to spend four beautiful years at IIIT Bhubaneswar where I worked with Prof. Anjali Mohapatra on Genetic Clustering Algorithms in Bioinformatics.

Interests. Differential Privacy, Generative AI, Data Cleaning, Databases

News

• [20.4.2026] I have joined as a senior scientist at CFAR, A*STAR Singapore. I am excited to work on privacy and safety of AI systems in this new role.
• [31.3.2026] I have been awarded the National Research Foundation Postdoctoral Award to conduct research on Private Learning and Safe Use of AI.
• [12.6.2026] I will be on the program committees of TPDP 2026 and SeQureDB 2026.
• [6.2.2026] Our work ‘CAPID: Context-Aware PII Detection for Question-Answering Systems’ has been accepted to EACL SRW 2026.
• [5.10.2025] I will be giving two talks on ‘End-to-End Private Learning: Challenges of Noisy and Inconsistent Data’ at UBC (8th Oct) and Vector Institute (9th Oct).
• [26.9.2025] I won the third place award at the poster competition at the Cheriton symposium 2025.
• [23.9.2025] Our work ‘MaskSQL: Safeguarding Privacy for LLM-Based Text-to-SQL via Abstraction’ has been accepted to Regulatable ML at NeurIPS 2025.
• [31.7.2025] I successfully defended my thesis titled ‘Towards practicable and effecient private learning’.
• [25.7.2025] I will be on the program committee of AAAI 2026.
• [17.3.2025] I will be giving a talk titled ‘End-to-End Private Learning: Challenges of Noisy and Inconsistent Data’ at INRIA for the PreMeDICaL team on April 1st.
• [26.1.2025] I will be giving two talks at IIT Madras on February 6 and IISc Bangalore on February 24.
• [31.01.2025] Our paper ‘Inconsistency measures for Differentially Private databases’ has been accepted to SIGMOD 2025.
• [10.10.2024] I am serving as a PC member for OnDBD 2024.
• [9.9.2024] I have started a new role as a research intern at Sony AI in the PPML team supervised by Vikash Sehwag.
• [4.5.2024] I received the Cheriton scholarship for CAD $10k per year for two years. It is given to top students at the Cheriton school of computer science on the basis of scholastic excellence.
• [13.4.2024] I am serving as a PC member for TPDP'24
• [15.3.2024] Differentially Private Data Generation with Missing Data is accepted to VLDB'24.

Selected Publications

For a full list of publications, please visit my Google Scholar profile.

• CAPID: Context-Aware PII Detection for Question-Answering Systems, EACL SRW 2026. [code] Mariia Ponomarenko, Sepideh Abedini, Masoumeh Shafieinejad, D.B.Emerson, Shubhankar Mohapatra, Xi He
• MaskSQL: Safeguarding Privacy for LLM-Based Text-to-SQL via Abstraction, NeurIPS 2025 (Regulatable ML Workshop). Sepideh Abedini, Shubhankar Mohapatra, D. B. Emerson, Masoumeh Shafieinejad, Jesse C. Cresswell, Xi He
• Inconsistency measures for Differentially Private databases, SIGMOD 2025 (also in TPDP 24). [code] [poster] [slides] Shubhankar Mohapatra, Amir Gilad, Xi He, Benny Kimelfeld
• Differentially Private Data Generation with Missing Data, VLDB 2024 (also in TPDP 22). [code] [poster] [slides] Shubhankar Mohapatra, Jianqiao Zong, Florian Kerschbaum, Xi He
• Rethinking Benchmarks for Private Image Classification, In Submission (also in TPDP 23). [code] Sabrina Mokhtari, Shubhankar Mohapatra, Sara Kodeiri, Florian Tramèr, Gautam Kamath
• The Role of Adaptive Optimizers for Honest Hyperparameter Tuning, AAAI 2022 (also in TPDP 21). [code] [poster] [slides] Shubhankar Mohapatra, Sajin Sasy, Xi He, Gautam Kamath, Om Thakkar
• Kamino: Constraint-Aware Differentially Private Data Synthesis, VLDB 2021. [code] Chang Ge, Shubhankar Mohapatra, Xi He, Ihab F. Ilyas

• TaxiTrip-Synthesizer, Differential Privacy Temporal Map Challenge (DeID2). [code] Christian Covington, Karl Knopf, Shubhankar Mohapatra Mohapatra, Shufan Zhang